“You did incredibly well!”: teachers’ inflated praise can make children from low-SES backgrounds seem less smart (but more hardworking)

The study examines whether teachers’ inflated praise—overly enthusiastic evaluations like “You did incredibly well!”—can inadvertently signal low ability and thus make children from low-SES backgrounds seem less smart. Achievement inequality persists even when ability is held constant, and teachers, aiming to help, often use praise. Building on attribution theory, the authors propose that teachers may be more inclined to attribute low-SES students’ success to effort (controllable, unstable) rather than ability (uncontrollable, stable), making success appear more praiseworthy and eliciting more inflated praise. Older children (from about age 10) understand that effort and ability can be compensatory; thus, seeing someone receive more praise may lead them to infer the praised student worked harder and is therefore less able. The authors hypothesize: (1) teachers provide more inflated (but not modest) praise to low-SES than to equally successful high-SES students because of effort attributions; and (2) children infer that students who receive inflated praise are less smart but more hardworking than equally performing peers who receive modest or no praise.

Prior work suggests teachers believe praise boosts motivation and self-esteem, yet praise varies in intensity and about a quarter is inflated. Attribution theory (locus, stability, controllability) distinguishes effort and ability as core causes of achievement, often seen as compensatory. Teachers’ motivational lenses bias reinforcement: low-ability but hardworking students may receive more rewards and fewer punishments. A positive feedback bias has been reported in ethnic/racial contexts: majority-group teachers sometimes provide more positive feedback to minority students and withhold criticism, occasionally delivering extremely positive comments. There is a pervasive stereotype that low-SES individuals have lower academic ability; even teachers underestimate low-SES children’s abilities. Developmentally, children older than ~10 infer that when two students achieve the same outcome, the one who worked harder is less smart, and prior studies show praised students can be judged as less smart than non-praised hardworking peers. However, effects of inflated praise on children’s inferences and SES-related biases in praise had not been directly tested. The present work integrates literatures on positive feedback bias, inflated praise, and attribution theory to examine SES-specific inflated praise and its interpretive consequences for children.

Design: Two preregistered experiments using vignettes ensured experimental control. Study 1 (teachers) used a within-subjects design to compare responses to equally successful high-SES and low-SES hypothetical students. Study 2 (children) used a within-subjects design to compare inferences about two equally performing peers who received different levels of teacher praise.

Study 1 (Teachers): Participants were 106 primary school teachers (ages 21–63) in the Netherlands/Belgium, recruited via professional channels; ethics approval obtained. Teachers read four randomized vignettes: two low-SES and two high-SES 11-year-old students, each achieving one of the highest grades in class. SES was cued via home/resources descriptions and illustrations; ability/outcome information was held constant. After each vignette, teachers wrote (a) what they would say to the student (open-ended) and (b) why the student performed well (open-ended). Responses were coded by two blind coders for praise (modest vs. inflated; inflated contained adverbs/adjectives like very, incredibly, amazing) and attributions (ability, effort, other). Multiple instances per response could be coded; inter-rater reliability was acceptable-to-good (praise total α≈0.73; modest α≈0.73; inflated α≈0.79; attributions ability α≈0.69; effort α≈0.63; other α≈0.70). Manipulation checks verified SES perception. Difference scores per teacher contrasted low-SES minus high-SES for total, modest, and inflated praise, and for attributions (ability, effort, other). Primary analyses used one-tailed Wilcoxon signed rank tests (α=0.05) per preregistration; exploratory analyses used two-tailed tests as noted.

Study 2 (Children): Participants were 63 primary school children (ages 10–13) tested during school hours in regular classrooms; target N=74 per power analysis (one-tailed z-tests, α=0.05, h=0.40) but recruitment ended at N=63 per preregistration; ethics approval and consent procedures followed. Children saw three randomized vignettes in which two students took the same exam and achieved identical scores (7/10), but teachers provided: (1) modest praise vs. no praise, (2) inflated praise vs. no praise, and (3) inflated praise vs. modest praise. After each vignette, children made forced-choice judgments: who is smarter, and who worked harder; they also provided open-ended reasons for why one student received (more) praise. Primary analyses tested whether the proportion judging the more-praised student as less smart exceeded 0.50 (one-tailed z-tests). Exploratory analyses tested effort inferences and moderation by gender, age, and subjective social status. Randomization covered praise assignment to left/right student and condition order.

Study 1 (Teachers):

• Total praise: No overall difference in total praise for low- vs. high-SES students (Wilcoxon signed rank W=87, p=0.644).
• Inflated praise: Significantly more inflated praise for low-SES than high-SES students (W=240, p<0.001).
• Modest praise: No preregistered one-tailed effect (W=71, p=0.997), but exploratory analysis indicated less modest praise for low-SES than high-SES students (W=71, p=0.003, one-tailed).
• Differential impact: The difference between inflated and modest praise difference scores was significantly >0 (V=465, p<0.001), indicating SES had a stronger effect on inflated than modest praise.
• Attributions (exploratory): Teachers were more inclined to attribute low-SES students’ success to effort (difference score location >0; V=338, p<0.001); ability attributions did not differ by SES (V=1, p=0.878); other attributions (often external causes) were less often used for low-SES than high-SES students (location <0; V=34, p<0.001).

Study 2 (Children):

• Ability inferences: Children judged praised students as less smart than non-praised peers when outcomes were equal. Specifically, more than half judged the modest-praised student as less smart than the non-praised student (z=5.71, p<0.001), and more than half judged the inflated-praised student as less smart than the non-praised student (z=2.484, p<0.001). Moreover, children judged the inflated-praised student as less smart than the modest-praised student (direction consistent with compensatory reasoning).
• Effort inferences (exploratory): Children inferred greater effort for praised students: modest praise vs. no praise (z=2.408, p<0.01), inflated praise vs. no praise (z=4.08, p<0.001), and inflated praise vs. modest praise (z=6.33, p<0.001).
• Moderation: No significant interactions with children’s gender, age, or subjective social status.

Overall: Teachers disproportionately gave inflated praise to low-SES students for identical success, and children interpreted inflated (and modest) praise as a cue of lower ability but higher effort when performance was equal.

The findings support the hypotheses that teachers’ inflated praise is more likely for low-SES students and that children interpret inflated praise as a signal of lower ability but greater effort. This pattern suggests teachers’ attributions—viewing low-SES students’ success as effort-driven—may underlie a positive feedback bias that unintentionally reinforces stereotypes about low-SES students’ intellectual ability. The work bridges attribution theory with research on inflated praise and feedback biases, highlighting that older children infer compensatory relations between effort and ability and thus draw ability-diminishing conclusions from inflated praise even when effort is not explicitly mentioned. The results imply that well-intentioned practices can backfire by shaping peer perceptions and potentially students’ self-views, thereby contributing to inequality. The authors discuss implications for practice: recruiting more teachers from low-SES backgrounds, structuring classrooms to emphasize learning over selection to sideline bias, and designing praise interventions that promote modest, specific, and equitable praise to avoid signaling low ability. They also note developmental, cross-cultural, and intersectional considerations and the importance of understanding how classroom feedback patterns may transmit intellectual stereotypes of low-SES groups.

Across two preregistered experiments, teachers gave more inflated praise to low-SES than to equally successful high-SES students, and children inferred that recipients of inflated praise were less smart but more hardworking. These results suggest that inflated praise can inadvertently reinforce negative stereotypes about the academic abilities of low-SES children. The paper contributes a mechanism—effort attributions—linking SES to feedback valence/intensity, integrates literatures on positive feedback bias and attribution, and underscores the need to refine praise practices. Future research should: (1) test how inflated praise shapes recipients’ self-views and subsequent achievement; (2) examine teachers’ responses to failures (e.g., pity, comfort-oriented feedback) across SES; (3) investigate cross-cultural variability in praise prevalence and meaning; and (4) consider intersectional identities to understand compounding biases in teacher feedback.

Key limitations include: (1) use of hypothetical vignettes to ensure control, which may heighten SES salience and reliance on stereotypes relative to naturalistic contexts; (2) acceptable but not excellent inter-rater reliability for coding praise/attributions and limited coder training materials; (3) focus on children aged 10–12, limiting developmental generalizability; (4) within-subjects design with similar vignettes could induce demand characteristics despite randomization; (5) Study 2 fell short of planned sample size (N=63 vs. 74), though preregistered stopping rules were followed; (6) online administration in Study 1 and possible motivational effects; and (7) SES operationalization through explicit material cues may not capture the full nuance of students’ backgrounds.